Machine Learning Lab

Possible Topics

Here, we collect possible topics for BSc/MSc projects or theses. Please note that the list of projects below is often incomplete, since we always have new ideas as time goes by and we do not always update the list right away.

Getting in Touch about Theses

Since machine learning is currently one of the hottest topics around, our small group is flooded with requests, and we may not be able to offer projects to every interested student. Generally, it helps for finding a matching project if you have taken and excelled at our courses. All projects in our research group should be carried out in teams of at least 2 students (max. 4). Theses can also be done in synergy with theses/projects by others, but each student has to write their thesis by themselves, appropriately acknowledging work done by others. We strongly recommend doing a MSc project with our group before inquiring about a MSc thesis.

Please see Theses in Industry below to see in which cases and how it is possible to do one.

Your application

In order to make the selection process of students and topics as effective as possible, we ask potential candidates for projects/theses to send us an email (optimally, already as a team) to ml-student-inquiry@cs.uni-freiburg.de (or only to the person(s) in charge of a specific project you're interested in) with the following information (per team member):

General

1. Are you a BSc or MSc student? Which term? If BSc, are you planning on staying in Freiburg for a MSc?
2. Which area of work do you prefer: Neural Architecture Search, Hyperparameter optimization, AutoML systems or Learning to Learn (L2L)?
3. Which ML-related courses have you taken?
4. Can you please attach your transcript of records?
5. Which courses are you taking which are NOT yet in the transcript of records?
6. Which projects have you done so far (in Freiburg and elsewhere)?
7. Which topics interest you most?

Your background

1. Creativity / ideas for developing new algorithms
2. Getting someone else's large code base to run
3. Running comprehensive experimental studies / keeping track of results
4. Self-motivation to push through even if things don't work for a while
5. Coding skills
   1. Python
   2. TensorFlow
   3. Keras
   4. PyTorch
   5. C/C++
6. Ability to read a RL paper, implement it and get it to work
7. Ability to read a DL paper, implement it and get it to work
8. Formal background, linear algebra
9. Formal background, proofs

Projects: Expected Time Frame and Outcome

The faculty does not set a hard deadline for projects as they do for theses (i.e., 6 months). However, to avoid projects running too long, our chair has decided to limit them to a maximum of 3 months of full-time work or 6 months of half-time work (deviations are possible, but should be agreed upon beforehand). The goal of a project with our group is to produce a short, yet high-quality research paper that could be submitted to a top machine learning conference workshop, such as those held at NeurIPS, ICML, or ICLR.

Requirements for Reproducible Experiments

When doing a project or thesis in our group, we put high value on reproducibility. This means that someone unfamiliar with your code can understand and reproduce what you have done, not only now, but also in the far future. In your final presentation, you will need to describe the steps you took to ensure reproducibility and this will affect your final grade. This minimally entails that all code you used must be made available and depending on the project, your supervisor may require you to containerize your code (e.g., Singularity). We strongly recommend you think about reproducibility at the very start of the project and not just before handing it in.

Please note: The research assistant jobs listed here are only available to students enrolled at the University of Freiburg.

Open Positions

Teaching Assistant for "Automated Machine Learning" (HiWi)

We are actively looking for good students that have successfully participated in the AutoML lecture, and that would like to assist us in organizing it in the future. This entails:

• maintaining old and creating new exercises for the course based on feedback from the previous semester (during the winter term)
• Help with grading exercises and answering questions related to the exercises (during the summer term)

You can start anytime during the winter term or with the start of a new summer semester.

Teaching Assistant for "Foundations of Deep Learning" (HiWi)

We are actively looking for good students that have taken and successfully completed the Foundations of Deep Learning lecture, and that would you like to assist us in organizing it in the future.

Teaching Assistant for "System Design Project" (HiWi)

We are actively looking for a student with strong web development skills for the "System Design Project". If you are interested please reach out to Frederic Runge.

Deep Learning Engineer (HiWi)

If you can quickly code algorithms from deep learning papers and repeat their experiments, we want you!

More information about these and other open positions can be found here.

Getting in Touch about Research Assistant Jobs

Since machine learning is currently one of the hottest topics around, our small group is flooded with requests, and we may not be able to offer projects to every interested student.

Your application

If you are interested in a specific project, please feel free to contact the project supervisor directly. Alternatively, if you are flexible and interested in multiple different projects, you may email to ml-student-inquiry@cs.uni-freiburg.de instead, clearly stating that you are applying for a HiWi job (and not a project/thesis). In both cases, we ask you to provide us with the following information:

General

1. Are you a BSc or MSc student? Which term? If BSc, are you planning on staying in Freiburg for a MSc?
2. Which area of work do you prefer: Neural Architecture Search, Hyperparameter optimization, AutoML systems or Learning to Learn (L2L)?
3. Which ML-related courses have you taken?
4. Can you please attach your transcript of records?
5. Which courses are you taking which are NOT yet in the transcript of records?
6. Which projects have you done so far (in Freiburg and elsewhere)?
7. Which topics / projects interest you most?

Your background

1. Creativity / ideas for developing new algorithms
2. Getting someone else's large code base to run
3. Running comprehensive experimental studies / keeping track of results
4. Self-motivation to push through even if things don't work for a while
5. Coding skills
   1. Python
   2. TensorFlow
   3. Keras
   4. PyTorch
   5. C/C++
6. Ability to read a RL paper, implement it and get it to work
7. Ability to read a DL paper, implement it and get it to work
8. Formal background, linear algebra
9. Formal background, proofs

Since different projects require different skill sets, please also rate your skills in the following categories on a scale from ++ (very good) to -- (no knowledge/skill):

Please state the number of hours per month you would like to work for.

If you are currently doing a postdoc or PhD in a field very closely related to our group’s research and would like to visit our group, please follow the procedure described on the open positions page.

If you do not have a MSc degree, there is no possibility for doing an internship with us. Rather, we encourage you to apply to our MSc programme. Information on this programme and how to apply can be found here: https://www.tf.uni-freiburg.de/en/study-programs/computer-science/m-sc-computer-science. We note that in the AI masters program you can basically fill your entire curriculum with AI courses.

DAAD Wise program: We do not accept applications for DAAD internships anymore.

We only supervise MSc theses in industry in the following exceptional circumstances:

• 1. The research topic is directly related to AutoML and discussed with us before the research starts.
• 2. The industry partner promises that all code (including scripts to run experiments, baselines, etc) you develop will be open source and freely available to the academic community
• 3. The industry partner promises that at least part of the evaluation is on publicly available data and therefore directly reproducible by the academic community
• 4. The industry partner promises that the results will be freely publishable and that there is no NDA we have to sign
• 5. The deliverables include the open-source code (2.) and a Docker or Singularity container that reproduces the results of the thesis on freely available data (3).

The promises referred to above are to be made in writing.